Would it be feasible to reduce the scope of the openfile lock when read or write file meta info?

[SeanHai]

func (m *baseMeta) Write(ctx Context, inode Ino, indx uint32, off uint32, slice Slice, mtime time.Time) syscall.Errno {
	defer m.timeit("Write", time.Now())
	f := m.of.find(inode)
	if f != nil {
		f.Lock()
		defer f.Unlock()
	}
	defer func() { m.of.InvalidateChunk(inode, indx) }()
	var numSlices int
	var delta dirStat
	var attr Attr
	st := m.en.doWrite(ctx, inode, indx, off, slice, mtime, &numSlices, &delta, &attr)
	if st == 0 {
		m.updateParentStat(ctx, inode, attr.Parent, delta.length, delta.space)
		if delta.space != 0 {
			m.updateUserGroupQuota(ctx, attr.Uid, attr.Gid, delta.space, 0)
		}
		if numSlices%100 == 99 || numSlices > 350 {
			if numSlices < maxSlices {
				go m.compactChunk(inode, indx, false, false)
			} else {
				m.compactChunk(inode, indx, true, false)
			}
		}
	}
	return st
}

What would you like to be added:
We found that during test 4k random writes on large file(10GiB) with fio tool, requests get blocked on the openfile lock(f.Lock) while writing metadata. Is it posssible reduce the whole openfile lock to the specific chunk?

Why is this needed:
There are lots of chunkWriters working in parallel, the data can be flush concurrently, but slice info must update in order. So there is a bottleneck when updating metadata.